feat(webdav): use Tika's merged thumbnail preset

Tika merged the thumbnail catalog preset and ThumbnailUnpackSelector
(TIKA-4856): PUT /unpack/preset/thumbnail returns exactly the one raster
image a client would show for the document - a stored thumbnail, the
rendering of a metafile thumbnail, or the rendered first page. The
client-side selection over /unpack/all (embedded-resource-type marks,
metafile renderings, id-path matching) moves to the server; the preset
zip carries one image without metadata sidecars, so the entry's
extension names the type and the bytes break the tie. A 404 now points
at the preset activation the server config needs.

services/webdav/README.md

@@ -10,7 +10,7 @@ Currently, the webdav service handles request for two functionalities, which are
 
 The webdav service provides various `GET` endpoints to get the thumbnails of a file in authenticated and unauthenticated contexts. It also provides thumbnails for spaces on different endpoints.
 
-Thumbnails for documents, raw photos, audio cover art and other formats the webdav service cannot render itself come from an Apache Tika server (4.1 or newer) when `OC_TIKA_URL` (or `WEBDAV_TIKA_URL`) is set. Every file that is not a directly supported image is sent to Tika; `OC_TIKA_THUMBNAIL_MIME_TYPES` (or `WEBDAV_TIKA_THUMBNAIL_MIME_TYPES`) restricts that to a list of mime types, where an entry like `image/x-raw-samsung:image/x-samsung-srw` maps a type to the one Tika knows the format by. Without Tika, audio cover art and GeoGebra thumbnails are still extracted by the webdav service itself; those built-in extractors are deprecated.
+Thumbnails for documents, raw photos, audio cover art and other formats the webdav service cannot render itself come from an Apache Tika server (4.1 or newer) when `OC_TIKA_URL` (or `WEBDAV_TIKA_URL`) is set. Every file that is not a directly supported image is sent to Tika; `OC_TIKA_THUMBNAIL_MIME_TYPES` (or `WEBDAV_TIKA_THUMBNAIL_MIME_TYPES`) restricts that to a list of mime types, where an entry like `image/x-raw-samsung:image/x-samsung-srw` maps a type to the one Tika knows the format by. The webdav service calls `PUT /unpack/preset/thumbnail`, so the Tika server must activate the `thumbnail` catalog preset in its configuration (`"presets": {"thumbnail": true}`). Without Tika, audio cover art and GeoGebra thumbnails are still extracted by the webdav service itself; those built-in extractors are deprecated.
 
 Generated thumbnails are cached by the webdav service itself. The cache backend defaults to `file`, storing entries under `$OC_BASE_DATA_PATH/thumbnails/files` (override with `WEBDAV_THUMBNAIL_CACHE_BACKEND` and `WEBDAV_THUMBNAIL_CACHE_DIR`). Use the `s3` backend when running multiple instances behind a load balancer so they share one cache.
 

services/webdav/pkg/preprocessor/tika.go

@@ -3,10 +3,11 @@ package preprocessor
 import (
 	"archive/zip"
 	"bytes"
-	"encoding/json"
 	"fmt"
 	"io"
+	"mime"
 	"net/http"
+	"path/filepath"
 	"strings"
 	"time"
 )
@@ -40,14 +41,17 @@ func (t TikaThumbnail) Convert(r io.Reader) (any, error) {
 	return Encoded{ContentType: contentType, Data: img}, nil
 }
 
-// tikaThumbnail returns the thumbnail Tika found in the document and its content
-// type. Tika 4.1 marks it as a THUMBNAIL embedded document, so /unpack/all is
-// enough: no preset, no endpoint of our own.
+// tikaThumbnail returns the thumbnail Tika found in the document and its
+// content type. The thumbnail catalog preset (Tika 4.1, TIKA-4856) unpacks
+// exactly the raster image a client would show for the document — a stored
+// thumbnail such as a raw photo's embedded preview or an audio file's cover,
+// the rendering of an office document's metafile thumbnail, or the rendered
+// first page of a PDF — so the zip carries that one image and nothing else.
 func tikaThumbnail(tikaURL, filename, contentType string, data []byte) (string, []byte, error) {
 	if contentType == "" {
 		contentType = "application/octet-stream"
 	}
-	req, err := http.NewRequest(http.MethodPut, strings.TrimRight(tikaURL, "/")+"/unpack/all", bytes.NewReader(data))
+	req, err := http.NewRequest(http.MethodPut, strings.TrimRight(tikaURL, "/")+"/unpack/preset/thumbnail", bytes.NewReader(data))
 	if err != nil {
 		return "", nil, err
 	}
@@ -68,6 +72,9 @@ func tikaThumbnail(tikaURL, filename, contentType string, data []byte) (string,
 	case http.StatusOK:
 	case http.StatusNoContent:
 		return "", nil, ErrNoThumbnail
+	case http.StatusNotFound:
+		// the route exists once the catalog preset is active
+		return "", nil, fmt.Errorf(`tika unpack returned %s: the "thumbnail" preset needs Tika >= 4.1 and "presets": {"thumbnail": true} in its config`, resp.Status)
 	default:
 		return "", nil, fmt.Errorf("tika unpack returned %s", resp.Status)
 	}
@@ -80,115 +87,32 @@ func tikaThumbnail(tikaURL, filename, contentType string, data []byte) (string,
 	if err != nil {
 		return "", nil, fmt.Errorf("tika unpack response: %w", err)
 	}
-	return thumbnailFromUnpack(zr)
+	return thumbnailFromZip(zr)
 }
 
-// vectorThumbnailTypes are thumbnail types nothing here can decode. Office
-// documents carry their preview as a metafile, so the THUMBNAIL entry is a
-// vector and the usable image is its rendering.
-var vectorThumbnailTypes = map[string]struct{}{
-	"image/emf":     {},
-	"image/wmf":     {},
-	"image/x-emf":   {},
-	"image/x-wmf":   {},
-	"image/svg+xml": {},
-}
-
-// unpacked is one entry of an /unpack/all response: the bytes and the metadata
-// are separate zip entries, "1.jpg" alongside "1.jpg.metadata.json".
-type unpacked struct {
-	name     string
-	kind     string
-	mimeType string
-	idPath   string
-}
-
-// thumbnailFromUnpack picks the image to use out of an unpack response. Tika
-// marks the document's preview as THUMBNAIL. For office documents that preview
-// is a metafile, and the raster we want is the RENDERING nested under it, which
-// tika only emits when its emf/wmf parser has renderImage turned on.
-func thumbnailFromUnpack(zr *zip.Reader) (string, []byte, error) {
-	var entries []unpacked
+// thumbnailFromZip takes the one image the preset unpacked. Lenient about the
+// shape: the first regular file that is not a metadata sidecar wins.
+func thumbnailFromZip(zr *zip.Reader) (string, []byte, error) {
 	for _, f := range zr.File {
-		if !strings.HasSuffix(f.Name, ".metadata.json") {
+		if f.FileInfo().IsDir() || strings.HasSuffix(f.Name, ".metadata.json") {
 			continue
 		}
-		meta, err := readZipJSON(f)
-		if err != nil {
-			continue
-		}
-		entries = append(entries, unpacked{
-			name:     strings.TrimSuffix(f.Name, ".metadata.json"),
-			kind:     metadataString(meta, "tk:embedded-resource-type"),
-			mimeType: metadataString(meta, "Content-Type"),
-			idPath:   metadataString(meta, "tk:embedded-id-path"),
-		})
-	}
-
-	for _, e := range entries {
-		if e.kind != "THUMBNAIL" {
-			continue
-		}
-		if _, vector := vectorThumbnailTypes[e.mimeType]; !vector {
-			return readThumbnail(zr, e)
-		}
-		if r, ok := renderingOf(entries, e); ok {
-			return readThumbnail(zr, r)
-		}
-		return "", nil, ErrNoThumbnail
-	}
-
-	// a document without a preview of its own can still have been rendered,
-	// a pdf page for instance
-	for _, e := range entries {
-		if e.kind == "RENDERING" {
-			return readThumbnail(zr, e)
+		img, err := readZipEntry(zr, f.Name)
+		if err != nil || len(img) == 0 {
+			return "", nil, ErrNoThumbnail
 		}
+		return entryContentType(f.Name, img), img, nil
 	}
 	return "", nil, ErrNoThumbnail
 }
 
-// renderingOf finds the rendering tika nested under a thumbnail. The rendering
-// is a child of the image it renders, so its id path extends the thumbnail's.
-func renderingOf(entries []unpacked, thumbnail unpacked) (unpacked, bool) {
-	for _, e := range entries {
-		if e.kind == "RENDERING" && thumbnail.idPath != "" && strings.HasPrefix(e.idPath, thumbnail.idPath+"/") {
-			return e, true
-		}
+// entryContentType names the image's type. The preset zip carries no metadata
+// entries, so the extension decides and the bytes break the tie.
+func entryContentType(name string, data []byte) string {
+	if byExt := mime.TypeByExtension(filepath.Ext(name)); byExt != "" {
+		return byExt
 	}
-	return unpacked{}, false
-}
-
-func readThumbnail(zr *zip.Reader, e unpacked) (string, []byte, error) {
-	img, err := readZipEntry(zr, e.name)
-	if err != nil || len(img) == 0 {
-		return "", nil, ErrNoThumbnail
-	}
-	return e.mimeType, img, nil
-}
-
-// readZipJSON decodes a metadata entry. Tika writes either an object or a
-// single element list.
-func readZipJSON(f *zip.File) (map[string]any, error) {
-	rc, err := f.Open()
-	if err != nil {
-		return nil, err
-	}
-	defer rc.Close()
-
-	var raw json.RawMessage
-	if err := json.NewDecoder(io.LimitReader(rc, maxTikaResponse)).Decode(&raw); err != nil {
-		return nil, err
-	}
-	var meta map[string]any
-	if err := json.Unmarshal(raw, &meta); err == nil {
-		return meta, nil
-	}
-	var list []map[string]any
-	if err := json.Unmarshal(raw, &list); err != nil || len(list) == 0 {
-		return nil, fmt.Errorf("unexpected metadata shape")
-	}
-	return list[0], nil
+	return http.DetectContentType(data)
 }
 
 func readZipEntry(zr *zip.Reader, name string) ([]byte, error) {
@@ -199,18 +123,3 @@ func readZipEntry(zr *zip.Reader, name string) ([]byte, error) {
 	defer rc.Close()
 	return io.ReadAll(io.LimitReader(rc, maxTikaResponse))
 }
-
-// metadataString reads a key; values are strings or lists.
-func metadataString(meta map[string]any, key string) string {
-	switch v := meta[key].(type) {
-	case string:
-		return v
-	case []any:
-		if len(v) > 0 {
-			if s, ok := v[0].(string); ok {
-				return s
-			}
-		}
-	}
-	return ""
-}

services/webdav/pkg/preprocessor/tika_test.go

@@ -3,7 +3,6 @@ package preprocessor
 import (
 	"archive/zip"
 	"bytes"
-	"encoding/json"
 	"image"
 	"image/color"
 	"image/png"
@@ -16,9 +15,9 @@ import (
 	"github.com/opencloud-eu/opencloud/services/webdav/pkg/thumbnail"
 )
 
-// unpackZip builds what /unpack/all returns: the entries and their metadata
-// side by side, "1.png" next to "1.png.metadata.json".
-func unpackZip(entries map[string][]byte, meta map[string]map[string]any) []byte {
+// presetZip builds what /unpack/preset/thumbnail returns: the one image the
+// preset selected, and nothing else.
+func presetZip(entries map[string][]byte) []byte {
 	var buf bytes.Buffer
 	zw := zip.NewWriter(&buf)
 	for name, data := range entries {
@@ -26,10 +25,6 @@ func unpackZip(entries map[string][]byte, meta map[string]map[string]any) []byte
 		Expect(err).ToNot(HaveOccurred())
 		_, err = w.Write(data)
 		Expect(err).ToNot(HaveOccurred())
-
-		w, err = zw.Create(name + ".metadata.json")
-		Expect(err).ToNot(HaveOccurred())
-		Expect(json.NewEncoder(w).Encode(meta[name])).To(Succeed())
 	}
 	Expect(zw.Close()).To(Succeed())
 	return buf.Bytes()
@@ -55,7 +50,7 @@ var _ = Describe("TikaThumbnail", func() {
 		thumbnail = func(w http.ResponseWriter, _ *http.Request) { w.WriteHeader(http.StatusNoContent) }
 		server = httptest.NewServer(http.HandlerFunc(func(w http.ResponseWriter, r *http.Request) {
 			requests = append(requests, r)
-			if r.URL.Path != "/unpack/all" {
+			if r.URL.Path != "/unpack/preset/thumbnail" {
 				w.WriteHeader(http.StatusNotFound)
 				return
 			}
@@ -67,19 +62,13 @@ var _ = Describe("TikaThumbnail", func() {
 		server.Close()
 	})
 
-	It("takes the entry Tika marked as the thumbnail, in one request", func() {
+	It("takes the image the preset unpacked, in one request", func() {
 		thumbnail = func(w http.ResponseWriter, r *http.Request) {
 			Expect(r.Method).To(Equal(http.MethodPut))
 			Expect(r.Header.Get("Content-Disposition")).To(ContainSubstring(`filename="shot.nef"`))
 			Expect(r.Header.Get("Content-Type")).To(Equal("image/x-nikon-nef"), "the file's type travels as the detection hint")
 			w.Header().Set("Content-Type", "application/zip")
-			_, _ = w.Write(unpackZip(
-				map[string][]byte{"0.nef": []byte("the raw itself"), "1.png": pngBytes()},
-				map[string]map[string]any{
-					"0.nef": {"Content-Type": "image/x-nikon-nef"},
-					"1.png": {"Content-Type": "image/png", "tk:embedded-resource-type": "THUMBNAIL"},
-				},
-			))
+			_, _ = w.Write(presetZip(map[string][]byte{"1.png": pngBytes()}))
 		}
 
 		img, err := TikaThumbnail{tikaURL: server.URL, filename: "shot.nef", contentType: "image/x-nikon-nef"}.Convert(bytes.NewReader([]byte("raw")))
@@ -88,81 +77,15 @@ var _ = Describe("TikaThumbnail", func() {
 		Expect(requests).To(HaveLen(1))
 	})
 
-	It("ignores embedded images that are not the thumbnail", func() {
+	It("decodes an image whose entry name has no telling extension", func() {
 		thumbnail = func(w http.ResponseWriter, _ *http.Request) {
 			w.Header().Set("Content-Type", "application/zip")
-			_, _ = w.Write(unpackZip(
-				map[string][]byte{"1.png": pngBytes()},
-				map[string]map[string]any{"1.png": {"Content-Type": "image/png", "tk:embedded-resource-type": "INLINE"}},
-			))
+			_, _ = w.Write(presetZip(map[string][]byte{"thumbnail": pngBytes()}))
 		}
 
-		_, err := TikaThumbnail{tikaURL: server.URL}.Convert(bytes.NewReader([]byte("raw")))
-		Expect(err).To(MatchError(ErrNoThumbnail), "an inline image is not a thumbnail, tika 4.0 marks cover art that way")
-	})
-
-	It("takes the rendering when the thumbnail is a metafile", func() {
-		thumbnail = func(w http.ResponseWriter, _ *http.Request) {
-			w.Header().Set("Content-Type", "application/zip")
-			_, _ = w.Write(unpackZip(
-				map[string][]byte{"1.wmf": []byte("vector"), "2.png": pngBytes()},
-				map[string]map[string]any{
-					"1.wmf": {"Content-Type": "image/wmf", "tk:embedded-resource-type": "THUMBNAIL", "tk:embedded-id-path": "/1"},
-					"2.png": {"Content-Type": "image/png", "tk:embedded-resource-type": "RENDERING", "tk:embedded-id-path": "/1/2"},
-				},
-			))
-		}
-
-		img, err := TikaThumbnail{tikaURL: server.URL, filename: "report.doc"}.Convert(bytes.NewReader([]byte("ole2")))
+		img, err := TikaThumbnail{tikaURL: server.URL}.Convert(bytes.NewReader([]byte("raw")))
 		Expect(err).ToNot(HaveOccurred())
-		Expect(img).ToNot(BeNil(), "office documents carry a metafile preview, the raster is its rendering")
-	})
-
-	It("reports no thumbnail when the metafile was not rendered", func() {
-		thumbnail = func(w http.ResponseWriter, _ *http.Request) {
-			w.Header().Set("Content-Type", "application/zip")
-			_, _ = w.Write(unpackZip(
-				map[string][]byte{"1.wmf": []byte("vector")},
-				map[string]map[string]any{
-					"1.wmf": {"Content-Type": "image/wmf", "tk:embedded-resource-type": "THUMBNAIL", "tk:embedded-id-path": "/1"},
-				},
-			))
-		}
-
-		_, err := TikaThumbnail{tikaURL: server.URL}.Convert(bytes.NewReader([]byte("ole2")))
-		Expect(err).To(MatchError(ErrNoThumbnail), "rendering is off in tika by default")
-	})
-
-	It("does not take a rendering that belongs to another image", func() {
-		thumbnail = func(w http.ResponseWriter, _ *http.Request) {
-			w.Header().Set("Content-Type", "application/zip")
-			_, _ = w.Write(unpackZip(
-				map[string][]byte{"1.wmf": []byte("vector"), "3.png": pngBytes()},
-				map[string]map[string]any{
-					"1.wmf": {"Content-Type": "image/wmf", "tk:embedded-resource-type": "THUMBNAIL", "tk:embedded-id-path": "/1"},
-					"3.png": {"Content-Type": "image/png", "tk:embedded-resource-type": "RENDERING", "tk:embedded-id-path": "/2/3"},
-				},
-			))
-		}
-
-		_, err := TikaThumbnail{tikaURL: server.URL}.Convert(bytes.NewReader([]byte("ole2")))
-		Expect(err).To(MatchError(ErrNoThumbnail))
-	})
-
-	It("takes a rendering of the document itself", func() {
-		thumbnail = func(w http.ResponseWriter, _ *http.Request) {
-			w.Header().Set("Content-Type", "application/zip")
-			_, _ = w.Write(unpackZip(
-				map[string][]byte{"1.png": pngBytes()},
-				map[string]map[string]any{
-					"1.png": {"Content-Type": "image/png", "tk:embedded-resource-type": "RENDERING", "tk:embedded-id-path": "/1"},
-				},
-			))
-		}
-
-		img, err := TikaThumbnail{tikaURL: server.URL, filename: "report.pdf"}.Convert(bytes.NewReader([]byte("pdf")))
-		Expect(err).ToNot(HaveOccurred())
-		Expect(img).ToNot(BeNil(), "a pdf has no preview of its own, a rendered page is one")
+		Expect(img).ToNot(BeNil(), "the bytes name the type when the entry does not")
 	})
 
 	It("reports a document without a thumbnail", func() {
@@ -172,6 +95,16 @@ var _ = Describe("TikaThumbnail", func() {
 		Expect(err).To(MatchError(ErrNoThumbnail))
 	})
 
+	It("reports an empty zip as no thumbnail", func() {
+		thumbnail = func(w http.ResponseWriter, _ *http.Request) {
+			w.Header().Set("Content-Type", "application/zip")
+			_, _ = w.Write(presetZip(nil))
+		}
+
+		_, err := TikaThumbnail{tikaURL: server.URL}.Convert(bytes.NewReader([]byte("zip")))
+		Expect(err).To(MatchError(ErrNoThumbnail))
+	})
+
 	It("reports a failing server", func() {
 		thumbnail = func(w http.ResponseWriter, _ *http.Request) { w.WriteHeader(http.StatusInternalServerError) }
 
@@ -179,11 +112,10 @@ var _ = Describe("TikaThumbnail", func() {
 		Expect(err).To(MatchError(ContainSubstring("500")))
 	})
 
-	It("reports a Tika that does not know the route", func() {
-		thumbnail = func(w http.ResponseWriter, _ *http.Request) { w.WriteHeader(http.StatusNotFound) }
-
-		_, err := TikaThumbnail{tikaURL: server.URL}.Convert(bytes.NewReader([]byte("raw")))
+	It("points at the preset config when the route is missing", func() {
+		_, err := TikaThumbnail{tikaURL: server.URL + "/nowhere"}.Convert(bytes.NewReader([]byte("raw")))
 		Expect(err).To(MatchError(ContainSubstring("404")))
+		Expect(err).To(MatchError(ContainSubstring("preset")), "a 404 means the catalog preset is not active")
 	})
 })